AgentMark Skills para Claude Code
agentmark-ai/skills
Skill para dominar AgentMark: CLI, prompts, datasets, evals e deploys
Ver no GitHubSkill que ensina agentes de IA como usar AgentMark — CLI, autoria de prompts, datasets, experimentos, evals e deploys baseados em git. Auto-gerada do repositório agentmark-ai/agentmark, mantém-se atualizada apontando para documentação viva.
Instalação
npx convez-ai agentmark-ai/skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Criar e iterar prompts estruturados com AgentMark
- Construir e validar datasets para treino de agentes
- Executar experimentos e avaliações automatizadas
- Fazer deploy de agentes via git com AgentMark
- Integrar AgentMark em workflows de Claude Code
Veredito

Skill bem arquitetada que funciona como ponte viva para AgentMark, delegando à documentação oficial em vez de replicar conteúdo estático. Ideal para quem quer automatizar ciclo completo de desenvolvimento, validação e deploy de agentes — evita debt de documentação ao apontar para fontes canônicas.
README
A tradução para o português deste README ainda está na fila.
AgentMark agent skills
Claude Code / agent skill that teaches AI agents how to use AgentMark — CLI commands, prompt authoring, datasets, experiments, evals, and git-based deploys.
Install
Pick whichever fits your tooling:
Universal (npx skills, cross-tool — Claude Code, Codex, Cursor, GitHub Copilot, Amp, Antigravity, +others):
npx skills add agentmark-ai/skills
Files land in ./.agents/skills/agentmark/ with symlinks into per-tool paths automatically.
Native Claude Code plugin marketplace:
claude
> /plugin marketplace add agentmark-ai/skills
> /plugin install agentmark
This uses Claude Code's built-in plugin system (.claude-plugin/marketplace.json in this repo declares the marketplace).
Mintlify well-known discovery (for any AgentMark docs URL):
npx skills add docs.agentmark.co
Once installed, Claude Code (or any agentskills.io-compatible tool) auto-discovers the skill at session start and invokes it when AgentMark context appears (.prompt.mdx files, agentmark.json, @agentmark-ai/* imports, or explicit mentions of AgentMark / CLI commands).
What's in this repo
skills/agentmark/
├── SKILL.md # entry point — read by Claude on auto-invoke
├── workflows/ # one file per common task
│ ├── creating-prompts.md
│ ├── building-datasets.md
│ ├── running-experiments.md
│ ├── using-evals.md
│ └── deploying.md
├── reference/ # auto-generated, do not hand-edit
│ ├── cli-commands.md # ← from cli-src/index.ts
│ ├── frontmatter-schema.md # ← from prompt-core schemas.ts
│ ├── api-commands.md # ← from openapi-spec.json
│ └── models.md # ← from model-registry
└── evals/scenarios.json # eval scenarios for skill validation
How it stays current
The skill is a thin pointer that delegates to live docs, not a static reference. It teaches agents to:
- Run
npx agentmark <cmd> --helpfor canonical CLI surface - Fetch
https://docs.agentmark.co/llms.txtfor doc navigation - WebFetch specific
.mdpages fromdocs.agentmark.cofor content
The four reference/*.md files are auto-generated from the canonical sources in agentmark-ai/agentmark on every release, so encoded facts can't drift.
Do not edit directly
This repo is auto-generated from agentmark-ai/agentmark.
Changes belong in skills/agentmark/ of the source repo. On every CLI release, a GitHub Action mirrors the source skills/ directory here. Direct commits to this repo will be overwritten on the next sync.
For substantive proposals, open an issue or PR against agentmark-ai/agentmark.
License
Same as AgentMark — AGPL-3.0-or-later.